It was back in the 1980s when Robert Freitas came up with a self-reproducing probe concept based on the British Interplanetary Society’s Project Daedalus, but extending it in completely new directions. Like Daedalus, Freitas’ REPRO probe would be fusion-based and would mine the atmosphere of Jupiter to acquire the necessary helium-3. Unlike Daedalus, REPRO would devote half its payload to what Freitas called its SEED package, which would use resources in a target solar system to produce a new REPRO probe every 500 years. Probes like this could spread through the galaxy over the course of a million years without further human intervention.
A Vision of Technological Propagation
I leave to wiser heads than mine the question of whether self-reproducing technologies like these will ever be feasible, or when. My thought is that I wouldn’t want to rule out the possibility for cultures significantly more advanced than ours, but the question is a lively one, as is the issue of whether artificial intelligence will ever take us to a ‘Singularity,’ beyond which robotic generations move in ways we cannot fathom. John Mathews discusses self-reproducing probes, as we saw yesterday, as natural extensions of our early planetary explorer craft, eventually being modified to carry out inspections of the vast array of objects in the Kuiper Belt and Oort Cloud.
Image: The Kuiper Belt and much larger Oort Cloud offer billions of targets for self-reproducing space probes, if we can figure out how to build them. Credit: Donald Yeoman/NASA/JPL.
Here is Mathews’ vision, operating under a System-of-Systems paradigm in which the many separate systems needed to make a self-reproducing probe (he calls them Explorer roBots, or EBs) are examined separately, and conceding that all of them must be functional for the EB to emerge (the approach thus includes not only the technological questions but also the ethical and economic issues involved in the production of such probes). Witness the probes in operation:
Once the 1st generation proto-EBs arrive in, say, the asteroid belt, they would evolve and manufacture the 2nd generation per the outline above. The 2nd generation proto-EBs would be launched outward toward appropriate asteroids and the Kuiper/Oort objects as determined by observations of the parent proto-EB and, as communication delays are relatively small, human/ET operators. A few generations of the proto-EBs would likely suffice to evolve and produce EBs capable of traversing interstellar distances either in a single “leap” or, more likely, by jumping from Oort Cloud to Oort Cloud. Again, it is clear that early generation proto-EBs would trail a communications network.
The data network — what Mathews calls the Explorer Network, or ENET — has clear SETI implications if you buy the idea that self-reproducing probes are not only possible (someday) but also likely to be how intelligent cultures explore the galaxy. Here the assumption is that extraterrestrials are likely, as we have been thus far, to be limited to speeds far below the speed of light, and in fact Mathews works with 0.01c as a baseline. If EBs are an economical and efficient way to exploring huge volumes of space, then the possibility of picking up the transmissions linking them into a network cannot be ruled out. Mathews envisages them building a library of their activities and knowledge gained that will eventually propagate back to the parent species.
A Celestial Network’s Detectability
Here we can give a nod to the existing work on extending Internet protocols into space, the intent being to connect remote space probes to each other, making the download of mission data far more efficient. Rather than pointing an enormous dish at each spacecraft in turn, we point at a spacecraft serving as the communications hub, downloading information from, say, landers and atmospheric explorers and orbiters in turn. Perhaps this early interplanetary networking is a precursor to the kind of networks that might one day communicate the findings of interstellar probes. Mathews notes the MESSENGER mission to Mercury, which has used a near-infrared laser ranging system to link the vehicle with the NASA Goddard Astronomical Observatory at a distance of 24 million kilometers (0.16 AU) as an example of what is feasible today.
Tomorrow’s ENET would be, in the author’s view, a tight-beam communications network. In SETI terms, such networks would be not beacons but highly directed communications, greatly compromising but not eliminating our ability to detect them. Self-reproducing probes propagating from star to star — conceivably with many stops along the way — would in his estimation use mm-wave or far-IR lasers, communicating through highly efficient and highly directive beams. From the paper:
The solar system and local galaxy is relatively unobscured at these wavelengths and so these signaling lasers would readily enable communications links spanning up to a few hundred AUs each. It is also clear that successive generations of EBs would establish a communications network forming multiple paths to each other and to “home” thus serving to update all generations on time scales small compared with physical transit times. These various generations of EBs would identify the locations of “nearby” EBs, establish links with them, and thus complete the communications net in all directions.
Working the math, Mathews finds that current technologies for laser communications yield reasonable photon counts out to the near edge of the Oort Cloud, given optimistic assumptions about receiver noise levels. It is enough, in any case, to indicate that future technologies will allow networked probes to communicate from one probe to another over time, eventually returning data to the source civilization. An extraterrestrial Explorer Network like this one thus becomes a SETI target, though not one whose wavelengths have received much SETI attention.
On Ethics and Possibilities
In any case, there is no reason why an exploring extraterrestrial culture would necessarily want its activities to be noticed. Rather than eavesdropping on leakage from an extremely efficient communications network, a more likely SETI outcome would involve human expansion through gradually more autonomous probes, with the chances of finding evidence for ET expanding as our own sphere of exploration widens. Getting a positive SETI result might thus involve centuries if not millennia.
It may also be the case that reproducing probes are severely restricted out of ethical concerns. Runaway propagation poses many dilemmas, so that few if any cultures build them. A null result might also indicate that their development is more difficult and expensive than anticipated, particularly in terms of finding needed energy sources.
How would we track narrow-beam communications systems in the mm-wave/IR region? As some commenters on yesterday’s post have already noted, they would likely be spread-spectrum, but there are tools for handling such signals. More on this, and on the ethics issues as well, tomorrow. Here again is the citation for the Mathews paper: “From Here to ET,” Journal of the British Interplanetary Society 64 (2011), pp. 234-241. For more on Robert Freitas’ REPRO ideas, see his paper “A Self-Reproducing Interstellar Probe,” JBIS 33 (July 1980), pp. 251-264.
It strikes me that when we’re talking about self-replicating probes that live off the energy and minerals found on-site to support reproduction and migration… that we’re discussing a microbe or similar single-cell carbon-like life pattern. Moderation of reproduction might take two tracks: 1) something like telemere truncation (is that the right term?) on DNA strands work to limit the number of times the cell can split, and 2) when population density hits some thresh-hold, children either need to move on (like spores scattered on interstellar winds), or specialize and change their role in the local environment (create coral beaches, or whatever).
In any event, the creator is left with the fundamental question – how to keep the created automatons from “finding the tree of life” and living forever.
The consequences of breaking the limited-lifetime rule is called cancer in today’s world. And are called Berserkers in Fred Saberhagen’s dystopian space opera.
How would a probe spawned in a new solar system locate other probes to set up a communications network? Each probe would be constructed in a new place and have no idea of what others might have been constructed “nearby,” whatever “nearby” would mean. Each probe would be utterly alone, spawned in the cold dark of space and having no knowledge of where its parents or siblings were, like an abandoned kitten in winter.
Would there be a failsafe or “KILL” signal? Imagine a network of probes growing unchecked… I know space is really big but perhaps you decide in advance how many generations you want to spawn. Generation 1 carries this number in memory. Say you desire 256 generations of probes. When Generation 2 is spawned it carries number 255 and subsequent generations get seeded with n-1 likewise. Problem is what happens when a cosmic ray hits the memory banks, flips a bit or two, and a probe in some remote generation gets its counter for example “199” turned into “E99.” (1×10^99) Then the entire universe fills up with these blasted things and before you know it, you can’t swing a dead cat without hitting a self replicating probe. This has serious privacy concerns associated with it.
So let’s say you build these things and before too long (million years or so) you have the entire galaxy populated with them, each phoning home over this vast galactic internet. You still have light time to contend with so best case you are still talking tens of thousands of years for the signal to get from the furthest reaches of the Milky Way. Will anyone be here to receive the data? What were we a million years ago? Tree dwelling Australopithecines? Cave dwelling Homo Erectusses? In a million years we might have brains so big that we can simply “think” our way across the galaxy. Or, more likely, we might have devolved back into something resembling the apes and when the signal comes in, we’ll simply pick an insect out of the fur of the ape next to us, pop it into our mouth, utterly oblivious to the far-infrared laser beam showering us with vast galactic knowledge.
-Mark
@Mark “In a million years we might have brains so big that we can simply “think” our way across the galaxy.”
In the interest of keeping an open mind in this curious world, it is worth mentioning that there are people who believe that they have already done that, in a manner of speaking:
“In Destiny of Souls, Dr. Michael Newton describes hybrid souls as those of “ mixed incarnation origins” who “have memories of actually incarnating on alien worlds …”
Ok, this is a crazy notion. But the people who assert that they have had these otherworldly experiences under hypnotic regression are no more obviously deranged than people who have faith that someday we will be able to build fusion powered probes which can self-replicate.
While skepticism about the subjective internal experiences of others is appropriate, these assertions are no more falsifiable than assertions about the prospects for von Neumann probes.
A Self-Reproducing Interstellar Probe
Robert A. Freitas Jr.
Journal of the British Interplanetary Society, Vol. 33, pp. 251-264 1980.
Note: This web version is derived from an earlier draft of the paper and may possibly differ in some substantial aspects from the final published paper.
Abstract:
Bracewell [1, 2] and Freitas [3] have discussed the possible superiority of interstellar probes in missions of galactic exploration and recently Calder [4] and Boyce [5] have raised the issue of self-organizing machines in related contexts.
In this paper a preliminary sketch of a self-reproducing starprobe is presented, with generation time ~103 years given a ~ 10-fold improvement in current human space/manufacturing technology.
http://www.rfreitas.com/Astro/ReproJBISJuly1980.htm
You are right, it is always very hard to falsify a “can do” statement. It is more easily proved correct: all you have to do is do the “do” part. So, the possibility of von Neumann probes will be proven when the first one is built. After one is not built in the next few centuries, I will concede that it might not be possible, although proof it is not.
I do not know how you get this idea.
My own intuition tells me that the new probe would be turned on inside a factory, surrounded by many other probes/robots, some just having been assembled, others still being assembled, yet others doing the assembling. The first thing the new probe would do is connect to the local database, which in turn is connected to the galactic internet. Here, all the information about anything is available as fast as possible under the light speed limitation. Including the mission and position of every other probe in the galaxy. Plus, of course, the new probe’s mission and destination, which will be promptly downloaded into its memory banks.
We do not have to content ourselves with a single number. The entire program under which the probes are deployed will be available. It has any needed limitations and failsafes built in as considered proper by the designers. There will be regular software updates from the designers and their descendants, who will likely be keeping a close eye on progress and keep making improvements, in software AND hardware.
There is no need to worry about bit flips, as all the information is stored under strong error correction codes, and distributed among the probes with enormous redundancy.
Will there still be designers around after a million years? I would optimistically think so, probably they will have by then followed in the footsteps of their machines to look at things with their own eyes. Or not. In any case, it is likely that the probes can continue their program forever even if any contact with intelligent beings ceases. Or, the designers could activate the failsafe procedures, which could destroy the system by turning everything off, or go into some sleep mode where replication is strictly at replacement level.
There is of course the possibility that the designers turn malicious, or a different group of malicious designers obtain the access codes, and the probes will be turned into the much-feared Berserkers of lore….
But it cannot happen accidentally, as long as the designers leave AI out of it.
“How would a probe spawned in a new solar system locate other probes to set up a communications network? Each probe would be constructed in a new place and have no idea of what others might have been constructed “nearby,” whatever “nearby” would mean. Each probe would be utterly alone, spawned in the cold dark of space and having no knowledge of where its parents or siblings were, like an abandoned kitten in winter. ”
What? No. Its parent is right next to it, having just built it. Its parent is also plugged into the interstellar commnet, and will plug the child into it. Its parent can know where siblings and cousins were and likely to have gone. Depending on how long it takes to replicate, it might be reasonable to keep pinging neighbors for updates, even through the nearby interstellar neighborhood.
One obvious restriction is “if you get to a new target and someone’s already there, self-terminate or move on without replication.”
Thanks for some really good articles on the von Neumann probes and The Fermi Paradox. From several exciting angles and ways of thinking. It is truly one of the most exciting topics in the community at the moment, which also gives a flood of articles on these topics. Really nice!
As much as I appreciate Freitas’ work he does make the dubious assumption in the REPRO study that the vehicle travels at the Daedalus top speed of ~0.12c. In order to brake REPRO has three stages and masses ~10.7 million tonnes. Halve the travel speed and the mass drops to just ~107,000 tonnes. Using the Daedalus pulse drive, with an exhaust velocity of ~0.035c, the 0.01c we’re discussing here would require a much lower fuel mass, so much so that the REPRO payload of ~900 tonnes, could be carried for just ~1160 tonnes of propellant. This makes the reproduction task that much easier. The question is just how much materials self-reproducing space vehicles would require per generation?
Since someone mentioned Chris Boyce, his argument was that even if there are vNMs in the Solar System, they might not consume terribly much. Assuming a 1,000,000 tonnes per vNM, a reproduction rate of ~1/10 year and an arrival rate of 1/10 year, then the total number of vNMs in our Solar System, over the last 4 billion years, is 1/2*(4E9/10)^2 ~8E+16. If all the “children” leave the solar system for other systems, then there’s ~4E8 somewhere in the outer system.
Adam, I am not sure I understand your (or Boyce’s?) calculation. Why would there be an arrival rate of 1/10 years? You would think that there would be no need for arrivals after the first. It also appears you (or he) tacitly assume there will be no exponential replication. This is of course possible, but then again, there could be, and then it would be a matter of a few millenia for all resources of the system to be consumed.
It is impossible, I think, to say anything about how much consumption there will be without knowing the intended mission. Physical laws allow for the entire range between total quiescence with just a few hidden facilities to wholesale consumption of the system in just a few dozen generations.
Hi Eniac
The assumption is that the vNMs aren’t mindless replicators, but intelligent, living beings, which are spreading themselves across the Galaxy to form a network. That they happen to be AI produced originally by organic intelligences doesn’t make vNMs any less intelligent or living. New arrivals are assumed to be from other intelligent species – 4E8 over 4E9 years isn’t overly many if we’re talking about civilizations arising at roughly the same pace as stars are born (which is what the Drake Equation boils down to by some analyses.)
A vNM arrives and proceeds to reproduce and explore its new star system. But only one seed-vehicle per star system, thus the daughter vNMs all depart for new systems. And the original doesn’t stop making daughters.
These are all reductio ad absurdum assumptions that Boyce is making, to argue against the Tiplerian assumption that whole solar systems would be consumed.
Using more reasonable assumptions then, if a seed vehicle made two daughters ever, then – assuming perfect survival – only ~(2^40) generations are needed to reach every star, brown dwarf and interstellar super-Jovian in the Galaxy. Another ~10 generations are required to visit every nomad object between the stars down to about ~Earth mass, assuming maximal numbers. If intelligent species arise and launch forth vNMs, roughly every decade on average, then over the last ~4 billion years, every independent object in the Galaxy is home to 400 million vNMs, studying it or now quiescent.
Reproduction is really what mortal entities need to do, but a being as capable and self-sufficient as a vNM might not feel so compelled to populate the Cosmos with copies of itself. Would networks inter-communicate amongst themselves, sharing information and the informational heritage of their originating species? I would expect so, since that’s one big reason to brave the void and span the Galaxy. Such a Galaxy-spanning informational web might be a truer picture of a Kardashev III meta-civilization – theoretically able to dam the flow of energy from the stars, but taking the long view that sidereal engineering might be needlessly grandiose and contaminating of their observational experiments.
“if a seed vehicle made two daughters ever, then – assuming perfect survival – only ~(2^40) generations are needed to reach every star, brown dwarf and interstellar super-Jovian in the Galaxy. ‘
I think you mean only 40 generations would be needed, generating 2^40 or about a trillion probes.
“intelligent living beings” isn’t a necessary assumption for vNMs. A certain amount of adaptability, yes, maybe a form of self-awareness, but not necessarily anything like human intelligence or free will, if you believe in free will.
Adam, thanks for the explanation, that makes sense now.
Of course, for me, it is hard to swallow the assumption that out of 4E8 probe networks every single one would be programmed (or evolve, or decide, as the case may be) to restrict itself to a single instance in each system and send away all of its progeny. While this may be a sensible approach for exploration, it is by no means a universal requirement. And what if all systems within propulsion range have been done, will the mother stop producing daughters? Why only settle nomads, why not also the proper planets, many or all of the asteroids, etc. etc?
Self-replicating probes for galactic exploration
By Antonia J. Jones – Department of Computing, Imperial College, London
http://users.cs.cf.ac.uk/O.F.Rana/Antonia.J.Jones/UnpublishedPapers/Probes.pdf